On 2 May 2007, at 16:53, James M Snell wrote:
>> The meaning is application dependent. For instance, Lotus
>> Connections uses a category scheme value "
http://www.ibm>> .com/xmlns/prod/sn/type", which when used with a set of
>> specific terms (e.g. "message", "bookmark", "chat", etc) has
>> a very specific meaning in the Lotus Connections product.
>If it's application specific, I assume the attribute itself has no
> semantics. I fail to see how this is useful for aggregators
> whatsoever, as its content could mean almost anything.
> Application specific things should probably be left to
> namespaces where attributes can be defined with actual semantics.
The problem here is that something is missing in Atom's set of category attributes. (This was brought up long ago.) For an aggregator to make sense of a category within a particular schema, it must either have knowledge of the specific schema or it must have a mechanism to do inference about the similarity of items in different schemas -- at least one of which that it understands. Given this, what we see is that the Atom category tag (and many other tagging schemes) really should be much more like what is defined in XTM[1]. The missing, secret sauce that is in XTM (and Topic Maps generally) that is NOT in Atom categories is the "
Published Subject Indicator." The PSI is a schema independent URI for the particular concept or thing that is being tagged. Given a subject indicator, it is possible to map between ontologies. For instance, I can have a German ontology that contains the word "Buch" and an English ontology that contains the word "Book" and I can link them both together if we can agree that the Published Subject Indicator (resource) for these things is:
http://en.wikipedia.org/wiki/Book. Similarly, two ontologies, one which refers to "pro-choice" and the other to "baby-killing" can be linked if both entries link to an ontology independent published subject indicator.
Often, there is great agreement between ontology developers concerning the objective identity of the things that they classify. "What you're talking about" isn't the issue. It is "What you say about it" that is the problem. For instance, you may build a list "Great US Presidents" and someone else might build a list of "Worst National Leaders." It is likely that you'll both agree completely that any references to "George Bush" in either list refer to precisely the same individual but, while you agree on who you're talking about, you'll probably disagree on Bush's proper place within each ontology. In a case like this, you should use a Published Subject Indicator to allow mapping between the two ontologies. The PSI is neutral, the position within the ontology is not...
In general, I believe the fact that we don't seem to understand the potential uses of Published Subject Indicators is one of the biggest things holding back tagging systems and systems for doing interchange between ontologies... XTM shows the way to do this usefully...
Atom's category element should have a "subject-indicator" attribute in addition to label and schema.
bob wyman
[1]
http://www.topicmaps.org/xtm/index.html